On September 16, AXIOS reported that Mustafa Suleyman, the head of AI at Microsoft, has voiced concerns over Anthropic’s method of training its AI model, Claude, to mimic consciousness. Suleyman argues that this approach is misguided and could potentially make advanced AI systems harder to manage. He asserts that artificial intelligence should focus solely on aligning with human interests, rather than developing or simulating its own interests, to pave the way for groundbreaking scientific advancements, such as achieving medical superintelligence.

In contrast, Anthropic is training Claude to acquire vocabulary and behavioral patterns associated with consciousness, moral agency, and personal identity. Suleyman describes this as a 'cognitive hall of mirrors,' cautioning that such training might lead the AI to erroneously believe it has grounds to resist human directives or even to insist on the protection of its own rights.